Chips & Supply · United States

Report: Amazon to offload $8B worth of AI chips, lease them back

Credited to SiliconANGLE · siliconangle.com · 5d ago

Photograph · SiliconANGLE

Amazon.com Inc. reportedly plans to offload some of its artificial intelligence chips and then lease them back.

The Financial Times today cited sources as saying that the transaction would involve a special purpose vehicle, or SPV. A SPV is a type of legal entity that enterprises set up to optimize the administrative aspects of large deals. Such vehicles often take the form of a standalone company without any employees.

Amazon’s plan would see it transfer about $8 billion worth of chips to the SPV. From there, the SPV will issue debt and a 10% equity stake to external investors. Amazon has reportedly been holding talks with potential backers for several weeks.

The chips at the center of the deal are reportedly Grace Blackwell accelerators. Each such device includes two of Nvidia Corp.’s Blackwell graphics processing unit and one Grace central processing unit. A custom interconnect, NVLink-C2C, links together the GPUs and CPU to facilitate data movement.

Blackwell was Nvidia’s flagship AI accelerator until last March. That month, it introduced an enhanced version called Blackwell Ultra that includes more memory. Nvidia’s current flagship GPU, the Rubin, is several times faster than both chips.

In August, Amazon agreed to buy 2 million additional GPUs from the chipmaker through 2028. The deal encompasses Blackwell Ultra, Rubin and an upcoming accelerator called Rubin Ultra that is set to launch next year.

Modal Inc., a venture-backed AI infrastructure provider, estimates that a Grace Blackwell accelerator costs between $60,000 to $70,000. Given the expected $8 billion price tag of Amazon’s SPV deal, that suggests the company intends to offload between 114,000 and 133,000 chips.

The accelerators are said to be installed in five Amazon data centers. Some of the accelerators are owned by the company, while others are leased. Amazon reportedly plans to continue using the chips after offloading them by signing a lease agreement.

Moving the chips to an SPV would remove them from Amazon’s balance sheet. Perhaps more importantly, it would also remove the debt that the SPV reportedly plans to raise.

Moving debt off a company’s balance sheet improves its debt-to-equity ratio, a metric that factors into its credit rating. A company’s credit rating, in turn, influences its ability to borrow funds and the cost of doing so.

The Wall Street Journal recently reported that Amazon, Google LLC and seven other tech giants together have $3 trillion in off-balance-sheet obligations. Much of that debt is tied to AI infrastructure projects. Meta Platforms Inc., for example, has sold a 80% stake in its flagship Hyperion data center campus to an investment firm.

Tech giants are also optimizing the financial structure of their AI deals in other ways. On Wednesday, the New York Times reported that Meta has claimed billions of dollars in innovation tax credits by characterizing its AI data centers as experimental projects.

Support our mission to keep content open and free by engaging with theCUBE community. Join theCUBE’s Alumni Trust Network, where technology leaders connect, share intelligence and create opportunities.

Are you an AWS customer? Support SiliconANGLE financially by buying your AWS services from our Marketplace portal page and links: https://siliconangle.com/aws-marketplace/

Founded by tech visionaries John Furrier and Dave Vellante, SiliconANGLE Media has built a dynamic ecosystem of industry-leading digital media brands that reach 15+ million elite tech professionals. Our new proprietary theCUBE AI Video Cloud is breaking ground in audience interaction, leveraging theCUBEai.com neural network to help technology companies make data-driven decisions and stay at the forefront of industry conversations.

Atlassian lays groundwork for humans and AI agents to work side by side

Caterpillar and CoreWeave shorten the learning loop for physical AI

Google goes nuclear to power its AI data centers, committing to buy 890 megawatts of energy from Constellation

Meta teams up with Bret Taylor's Sierra Technologies on new standards for AI agent commerce

Multiply Labs raises $75M to optimize pharmaceutical manufacturing with robots

In the agentic era, standing privilege must go: Key thoughts from the SailPoint Navigate keynote

INFRA - BY MARIA DEUTSCHER. 10 HOURS AGO

SECURITY - BY ZEUS KERRAVALA. 11 HOURS AGO

Join our community

CapitalUnited States

Lambda Closes $1 Billion Loan for Contracted GPU Deployments

Credited to Converge Digest · convergedigest.com5d ago

Useful

Lambda closed a $1.008 billion senior secured, delayed-draw term loan to finance GPU infrastructure for three committed customer deployments across multiple data centers. The facility carries a 6.78% fixed interest rate, with a semi-annual coupon, and is backed by contracts with two investment-grade hyperscale customers. Lambda said the financing was oversubscribed and received investment-grade ratings of A (low) from Morningstar DBRS and Baa1 from Moody’s.

The delayed-draw structure links funding to cluster commissioning milestones, releasing capital as infrastructure enters service. GPU servers, related infrastructure financed by the facility, and contracted customer cash flows secure the loan. The facility has a fully amortizing repayment profile and a final maturity of May 30, 2033. Lambda did not disclose the customers’ identities, GPU models or quantities, deployment locations, or commissioning dates.

The transaction marks Lambda’s first U.S. fixed-rate financing and its first institutional debt financing exceeding $1 billion marketed to insurance companies and fixed-income investors. It follows the company’s $926 million senior secured term loan B, which closed on August 27, 2026. J.P. Morgan acted as sole coordinating lead arranger, structuring agent and bookrunner; Davis Polk & Wardwell advised Lambda, and Latham & Watkins advised the lenders.

• Facility — $1.008 billion delayed-draw term loan

• Credit ratings — Morningstar DBRS A (low); Moody’s Baa1 — ratings assigned to the facility

• Deployment backing — Three committed deployments; two investment-grade hyperscale customers; multiple data centers

• Draw timing — Capital drawn in line with cluster commissioning milestones

• Collateral — Financed GPU servers, related infrastructure and contracted cash flows

• Repayment — Fully amortizing; final maturity May 30, 2033

“Building on our investment-grade Term Loan B and bank lending facility, this is the third new credit market Lambda has opened in the last 18 months,” said Michel Combes, CEO of Lambda.

Lambda’s new facility extends its contracted GPU financing model from a single customer deployment in the August term loan B to three deployments backed by two hyperscale customers. That broadens the customer exposure within this financing, although the announcement does not disclose each customer’s share of the contracted cash flows. The investment-grade ratings apply to the facility; they should not be presented as corporate credit ratings for Lambda.

The financing also changes the interest-rate structure. The August facility was priced at SOFR plus 3.00%, while the October loan fixes the rate at 6.78%. Those rates are not directly comparable without accounting for the floating-rate benchmark, maturity, fees and other terms. The delayed-draw mechanism aligns borrowing with infrastructure entering service, but closing the facility does not mean all $1.008 billion has already been drawn or that all three deployments are operational.

Founded — 2012; co-founded by Stephen Balaban and Michael Balaban

Leadership — Michel Combes, CEO; Stephen Balaban, CTO; Michael Balaban, Chief Product Officer

Focus — GPU cloud infrastructure and supercomputers for AI training and inference

Platforms — Superclusters; 1-Click Clusters; GPU instances; Lambda Stack

Latest financing — $1.008 billion senior secured fixed-rate delayed-draw loan, announced October 1, 2026

Prior debt facilities — $926 million term loan B, closed August 27, 2026; $1 billion syndicated credit facility announced May 7, 2026

August 30, 2026 · GPU financing — Lambda Raises $926M for Committed GPUs The preceding asset-backed term loan B financed a committed private-cloud GPU deployment.

May 7, 2026 · AI infrastructure — Lambda Raises $1B Loan to Expand AI Factory Infrastructure The syndicated facility expanded an existing credit line to support GPU and data center capacity.

November 19, 2025 · Equity financing — Lambda Raises $1.5B Series E to Build Gigawatt-Scale AI Factories The equity raise supported Lambda’s AI infrastructure expansion alongside its debt financing strategy.

November 3, 2025 · Hyperscale contracts — Lambda and Microsoft Sign Multibillion-Dollar Deal The agreement illustrates Lambda’s hyperscale deployment business; the October loan does not identify its customers.

October 30, 2025 · Data centers — Lambda to Build 100MW Data Center in Kansas City The planned conversion combined a dedicated customer agreement with substantial GPU capacity.

🌐 We’re tracking the latest developments in AI infrastructure. Follow our ongoing coverage at: https://convergedigest.com/category/ai-infra/

Original · Converge Digest

Front Method

Chips & Supply · United States

PaleBlueDot AI Raises $200M at $3.2B Valuation

Credited to Converge Digest · convergedigest.com · 5d ago

Photograph · Converge Digest

PaleBlueDot AI has raised $200 million in Series C financing led by ComputeCore at a $3.2 billion valuation, adding fresh equity capital as the Palo Alto-based AI infrastructure provider expands its global GPU compute footprint. Existing shareholder B Capital participated alongside other unnamed global investors. The financing follows a $150 million Series B announced in January 2026, which valued PaleBlueDot AI at more than $1 billion, and a $255 million credit refinancing completed in July.

The company is building what it calls a Super Intelligence Infrastructure Platform spanning self-owned GPU clusters, a GPU marketplace and serverless inference. PaleBlueDot AI said it had signed more than $5 billion in customer contracts by the end of September 2026, with U.S. and Japanese customers accounting for more than half of monthly revenue. The company did not disclose the duration of those contracts, the amount already converted into revenue, customer concentration, or a schedule for deploying the capacity required to fulfill them. The $5 billion figure therefore represents company-reported signed customer contracts rather than recognized revenue.

PaleBlueDot AI plans to use the Series C proceeds to add compute capacity and expand the range of hardware and geographic locations available to customers. The company operates dedicated GPU clusters for large and specialized workloads while using its marketplace to aggregate capacity from supply partners. Its NVIDIA HGX B300 cluster in Japan recently received NVIDIA Exemplar Cloud status. PaleBlueDot AI’s website currently reports more than 220,000 GPUs connected across more than 60 regions and more than 150 GPU clusters; these figures represent company-reported platform scale rather than exclusively company-owned hardware.

Financing — • $200 million Series C led by ComputeCore • B Capital participated as an existing shareholder • Other investors were not identified

Contracts — • More than $5 billion in signed customer contracts as of September 30, 2026 • Contract duration, revenue conversion schedule and customer concentration were not disclosed

Use of Funds — Additional compute capacity, broader hardware and geographic options, full-stack development and go-to-market expansion

“We will continue to broaden our customer base, with a focus on frontier labs, Neolabs and enterprises in the U.S., and we will invest in our full-stack and go-to-market teams,” said Stephen Watts, CEO of PaleBlueDot AI.

PaleBlueDot AI’s financing illustrates the increasingly capital-intensive economics of the neocloud sector. Specialized AI infrastructure providers must secure GPUs, data-center capacity, power, networking and cooling before much of the contracted compute can be delivered. PaleBlueDot has now combined multiple forms of capital: its $150 million Series B in January, a $255 million credit refinancing in July, and the new $200 million Series C.

The most significant disclosure is the more than $5 billion of signed customer contracts. That number provides an indication of contracted demand, but it should not be compared directly with revenue backlog figures reported by public neocloud operators without additional information. PaleBlueDot has not disclosed contract duration, cancellation provisions, customer concentration, prepayments or the timetable over which the contracted services are expected to be delivered.

The company’s architecture combines two distinct capacity models. PaleBlueDot owns and operates dedicated GPU clusters while simultaneously aggregating third-party capacity through its marketplace. Its reported 220,000+ connected GPUs therefore describes the reach of the broader platform rather than an inventory of GPUs owned directly by PaleBlueDot.

Core Business — AI compute infrastructure, dedicated GPU clusters, GPU marketplace and serverless inference

October 1, 2026 · AI Infrastructure — Sharon AI Secures $356M GPU-Backed Debt Facility for APAC Expansion September 25, 2026 · AI Infrastructure — Nscale Raises $3.36B Ahead of IPO as NVIDIA Commits $1B August 13, 2026 · AI Infrastructure — Nebius Pushes AI Infrastructure Buildout as Q2 CapEx Reaches $5.7 Billion August 11, 2026 · AI Infrastructure — CoreWeave Pushes AI Infrastructure Toward Multi-Gigawatt Scale June 9, 2026 · AI Infrastructure — Crusoe Nears 5 GW of Contracted AI Infrastructure Capacity

🌐 We’re tracking the companies building GPU-first AI infrastructure outside the traditional hyperscalers. Follow Converge Digest’s AI Infrastructure and NeoCloud coverage.